In recent years, 2D convolutional neural networks (CNNs) have been extensively used to diagnose neurological diseases from magnetic resonance imaging (MRI) data due to their potential to discern ...subtle and intricate patterns. Despite the high performances reported in numerous studies, developing CNN models with good generalization abilities is still a challenging task due to possible data leakage introduced during cross-validation (CV). In this study, we quantitatively assessed the effect of a data leakage caused by 3D MRI data splitting based on a 2D slice-level using three 2D CNN models to classify patients with Alzheimer's disease (AD) and Parkinson's disease (PD). Our experiments showed that slice-level CV erroneously boosted the average slice level accuracy on the test set by 30% on Open Access Series of Imaging Studies (OASIS), 29% on Alzheimer's Disease Neuroimaging Initiative (ADNI), 48% on Parkinson's Progression Markers Initiative (PPMI) and 55% on a local de-novo PD Versilia dataset. Further tests on a randomly labeled OASIS-derived dataset produced about 96% of (erroneous) accuracy (slice-level split) and 50% accuracy (subject-level split), as expected from a randomized experiment. Overall, the extent of the effect of an erroneous slice-based CV is severe, especially for small datasets.
Schizophrenia is among the most debilitating neuropsychiatric disorders. However, clear neurophysiological markers that would identify at-risk individuals represent still an unknown. The aim of this ...study was to investigate possible alterations in the resting alpha oscillatory activity in normal population high on schizotypy trait, a physiological condition known to be severely altered in patients with schizophrenia. Direct comparison of resting-state EEG oscillatory activity between Low and High Schizotypy Group (LSG and HSG) has revealed a clear right hemisphere alteration in alpha activity of the HSG. Specifically, HSG shows a significant slowing down of right hemisphere posterior alpha frequency and an altered distribution of its amplitude, with a tendency towards a reduction in the right hemisphere in comparison to LSG. Furthermore, altered and reduced connectivity in the right fronto-parietal network within the alpha range was found in the HSG. Crucially, a trained pattern classifier based on these indices of alpha activity was able to successfully differentiate HSG from LSG on tested participants further confirming the specific importance of right hemispheric alpha activity and intrahemispheric functional connectivity. By combining alpha activity and connectivity measures with a machine learning predictive model optimized in a nested stratified cross-validation loop, current research offers a promising clinical tool able to identify individuals at-risk of developing psychosis (i.e., high schizotypy individuals).
Physical inactivity has been identified as an important risk factor for dementia. High levels of cardiorespiratory fitness (CRF) have been shown to reduce the risk of dementia. However, the mechanism ...by which exercise affects brain health is still debated. Fractal dimension (FD) is an index that quantifies the structural complexity of the brain. The purpose of this study was to investigate the effects of a 5-year exercise intervention on the structural complexity of the brain, measured through the FD, in a subset of 105 healthy older adults participating in the randomized controlled trial Generation 100 Study. The subjects were randomized into control, moderate intensity continuous training, and high intensity interval training groups. Both brain MRI and CRF were acquired at baseline and at 1-, 3- and 5-years follow-ups. Cortical thickness and volume data were extracted with FreeSurfer, and FD of the cortical lobes, cerebral and cerebellar gray and white matter were computed. CRF was measured as peak oxygen uptake (VO2peak) using ergospirometry during graded maximal exercise testing. Linear mixed models were used to investigate exercise group differences and possible CRF effects on the brain's structural complexity. Associations between change over time in CRF and FD were performed if there was a significant association between CRF and FD. There were no effects of group membership on the structural complexity. However, we found a positive association between CRF and the cerebral gray matter FD (p < 0.001) and the temporal lobe gray matter FD (p < 0.001). This effect was not present for cortical thickness, suggesting that FD is a more sensitive index of structural changes. The change over time in CRF was associated with the change in temporal lobe gray matter FD from baseline to 5-year follow-up (p < 0.05). No association of the change was found between CRF and cerebral gray matter FD. These results demonstrated that entering old age with high and preserved CRF levels protected against loss of structural complexity in areas sensitive to aging and age-related pathology.
Abstract This study aims to assess the predictive capability of cylindrical Tumor Growth Rate (cTGR) in the prediction of early progression of well-differentiated gastro-entero-pancreatic tumours ...after Radio Ligand Therapy (RLT), compared to the conventional TGR. Fifty-eight patients were included and three CT scans per patient were collected at baseline, during RLT, and follow-up. RLT response, evaluated at follow-up according to RECIST 1.1, was calculated as a percentage variation of lesion diameters over time (continuous values) and as four different RECIST classes. TGR between baseline and interim CT was computed using both conventional (approximating lesion volume to a sphere) and cylindrical (called cTGR, approximating lesion volume to an elliptical cylinder) formulations. Receiver Operating Characteristic (ROC) curves were employed for Progressive Disease class prediction, revealing that cTGR outperformed conventional TGR (area under the ROC equal to 1.00 and 0.92, respectively). Multivariate analysis confirmed the superiority of cTGR in predicting continuous RLT response, with a higher coefficient for cTGR (1.56) compared to the conventional one (1.45). This study serves as a proof of concept, paving the way for future clinical trials to incorporate cTGR as a valuable tool for assessing RLT response.
Prostate cancer (PCa) is the most frequent male malignancy and the assessment of PCa aggressiveness, for which a biopsy is required, is fundamental for patient management. Currently, multiparametric ...(mp) MRI is strongly recommended before biopsy. Quantitative assessment of mpMRI might provide the radiologist with an objective and noninvasive tool for supporting the decision-making in clinical practice and decreasing intra- and inter-reader variability. In this view, high dimensional radiomics features and Machine Learning (ML) techniques, along with Deep Learning (DL) methods working on raw images directly, could assist the radiologist in the clinical workflow. The aim of this study was to develop and validate ML/DL frameworks on mpMRI data to characterize PCas according to their aggressiveness. We optimized several ML/DL frameworks on T2w, ADC and T2w+ADC data, using a patient-based nested validation scheme. The dataset was composed of 112 patients (132 peripheral lesions with Prostate Imaging Reporting and Data System (PI-RADS) score ≥ 3) acquired following both PI-RADS 2.0 and 2.1 guidelines. Firstly, ML/DL frameworks trained and validated on PI-RADS 2.0 data were tested on both PI-RADS 2.0 and 2.1 data. Then, we trained, validated and tested ML/DL frameworks on a multi PI-RADS dataset. We reported the performances in terms of Area Under the Receiver Operating curve (AUROC), specificity and sensitivity. The ML/DL frameworks trained on T2w data achieved the overall best performance. Notably, ML and DL frameworks trained and validated on PI-RADS 2.0 data obtained median AUROC values equal to 0.750 and 0.875, respectively, on unseen PI-RADS 2.0 test set. Similarly, ML/DL frameworks trained and validated on multi PI-RADS T2w data showed median AUROC values equal to 0.795 and 0.750, respectively, on unseen multi PI-RADS test set. Conversely, all the ML/DL frameworks trained and validated on PI-RADS 2.0 data, achieved AUROC values no better than the chance level when tested on PI-RADS 2.1 data. Both ML/DL techniques applied on mpMRI seem to be a valid aid in predicting PCa aggressiveness. In particular, ML/DL frameworks fed with T2w images data (objective, fast and non-invasive) show good performances and might support decision-making in patient diagnostic and therapeutic management, reducing intra- and inter-reader variability.
Pooling publicly-available MRI data from multiple sites allows to assemble extensive groups of subjects, increase statistical power, and promote data reuse with machine learning techniques. The ...harmonization of multicenter data is necessary to reduce the confounding effect associated with non-biological sources of variability in the data. However, when applied to the entire dataset before machine learning, the harmonization leads to data leakage, because information outside the training set may affect model building, and potentially falsely overestimate performance. We propose a 1) measurement of the efficacy of data harmonization; 2) harmonizer transformer, i.e., an implementation of the ComBat harmonization allowing its encapsulation among the preprocessing steps of a machine learning pipeline, avoiding data leakage by design. We tested these tools using brain T
-weighted MRI data from 1740 healthy subjects acquired at 36 sites. After harmonization, the site effect was removed or reduced, and we showed the data leakage effect in predicting individual age from MRI data, highlighting that introducing the harmonizer transformer into a machine learning pipeline allows for avoiding data leakage by design.
Abstract
Radiomics is emerging as a promising and useful tool in cardiac magnetic resonance (CMR) imaging applications. Accordingly, the purpose of this study was to investigate, for the first time, ...the effect of image resampling/discretization and filtering on radiomic features estimation from quantitative CMR T1 and T2 mapping. Specifically, T1 and T2 maps of 26 patients with hypertrophic cardiomyopathy (HCM) were used to estimate 98 radiomic features for 7 different resampling voxel sizes (at fixed bin width), 9 different bin widths (at fixed resampling voxel size), and 7 different spatial filters (at fixed resampling voxel size/bin width). While we found a remarkable dependence of myocardial radiomic features from T1 and T2 mapping on image filters, many radiomic features showed a limited sensitivity to resampling voxel size/bin width, in terms of intraclass correlation coefficient (> 0.75) and coefficient of variation (< 30%). The estimate of most textural radiomic features showed a linear significant (p < 0.05) correlation with resampling voxel size/bin width. Overall, radiomic features from T2 maps have proven to be less sensitive to image preprocessing than those from T1 maps, especially when varying bin width. Our results might corroborate the potential of radiomics from T1/T2 mapping in HCM and hopefully in other myocardial diseases.
Background
The relative contribution of changes in the cerebral white matter (WM) and cortical gray matter (GM) to the transition to dementia in patients with mild cognitive impairment (MCI) is not ...yet established. In this longitudinal study, we aimed to analyze MRI features that may predict the transition to dementia in patients with MCI and T
2
hyperintensities in the cerebral WM, also known as leukoaraiosis.
Methods
Sixty-four participants with MCI and moderate to severe leukoaraiosis underwent baseline MRI examinations and annual neuropsychological testing over a 2 year period. The diagnosis of dementia was based on established criteria. We evaluated demographic, neuropsychological, and several MRI features at baseline as predictors of the clinical transition. The MRI features included visually assessed MRI features, such as the number of lacunes, microbleeds, and dilated perivascular spaces, and quantitative MRI features, such as volumes of the cortical GM, hippocampus, T
2
hyperintensities, and diffusion indices of the cerebral WM. Additionally, we examined advanced quantitative features such as the fractal dimension (FD) of cortical GM and WM, which represents an index of tissue structural complexity derived from 3D-T
1
weighted images. To assess the prediction of transition to dementia, we employed an XGBoost-based machine learning system using SHapley Additive exPlanations (SHAP) values to provide explainability to the machine learning model.
Results
After 2 years, 18 (28.1%) participants had transitioned from MCI to dementia. The area under the receiving operator characteristic curve was 0.69 (0.53, 0.85) mean (90% confidence interval). The cortical GM-FD emerged as the top-ranking predictive feature of transition. Furthermore, aggregated quantitative neuroimaging features outperformed visually assessed MRI features in predicting conversion to dementia.
Discussion
Our findings confirm the complementary roles of cortical GM and WM changes as underlying factors in the development of dementia in subjects with MCI and leukoaraiosis. FD appears to be a biomarker potentially more sensitive than other brain features.